Section: Overall Objectives

Overall objectives

DIVERSE's research agenda targets core values of software engineering. In this fundamental domain we focus and develop models, methodologies and theories to address major challenges raised by the emergence of several forms of diversity in the design, deployment and evolution of software-intensive systems. Software diversity has emerged as an essential phenomenon in all application domains born by our industrial partners. These application domains range from complex systems brought by systems of systems (addressed in collaboration with Thales, Safran, CEA and DGA) and Instrumentation and Control (addressed with EDF) to pervasive combinations of Internet of Things and Internet of Services (addressed with TellU and Orange) and tactical information systems (addressed in collaboration with civil security). Today these systems seem to be radically different, but we envision a strong convergence of the scientific principles that underpin their construction and validation, bringing forwards sane and reliable methods for the design of flexible and open yet dependable systems. Flexibility and openness are critical and challenging software layer properties that must deal with four dimensions of diversity: diversity of languages, used by the stakeholders involved in the construction of these systems; diversity of features, required by the different customers; diversity of runtime environments, in which software has to run and adapt; diversity of implementations, which are necessary for resilience by redundancy.

In this context, the central software engineering challenge consists in handling diversity from variability in requirements and design to heterogeneous and dynamic execution environments. In particular this requires considering that the software system must adapt, in unpredictable yet valid ways, to changes in the requirements and environment. Conversely, explicitly handling of diversity is a great opportunity to allow software to spontaneously explore alternative design solutions. Concretely, we want to provide software engineers with the ability:

• to characterize an `envelope' of possible variations;

• to compose `envelopes' (to discover new macro envelopes in an opportunistic manner);

• to dynamically synthesize software inside a given envelop.

The major scientific objective that we must achieve to provide such mechanisms for software engineering is summarized below:

Scientific objective for DIVERSE: To automatically compose and synthesize software diversity from design to runtime to address unpredictable evolutions of software-intensive systems

Software product lines and associated variability modeling formalisms represent an essential aspect of software diversity, which we already explored in the past, and this aspect stands as a major foundation of DIVERSE's research agenda. However, DIVERSE also exploits other foundations to handle new forms of diversity: type theory and models of computation for the composition of languages; distributed algorithms and pervasive computation to handle the diversity of execution platforms; functional and qualitative randomized transformations to synthesize diversity for robust systems.